Learning low-rank latent mesoscale structures in networks